You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Karl Fogel <kf...@google.com> on 2006/10/03 21:18:29 UTC
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
rooneg@tigris.org writes:
> +What's wrong with this? Well, what if we need to rename A/B to X/Y so that
> +we can later rename X/Y to A/B? Not sure if there's a work around for that,
> +but I'm also not sure if there's a way to get that into the system anyway.
I think when something like that happens in a single commit txn, we'd
just "fold it down" (or "reduce to lowest common denominator", to use
one more metaphor before actually saying exactly what I mean).
In other words,
mv A/B X/Y
mv X/Y A/B
should reduce to a no-op. Do we actually do this folding right now?
I'm not sure; I think we should, though.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On 10/3/06, Mark Phippard <ma...@softlanding.com> wrote:
> Karl Fogel <kf...@google.com> wrote on 10/03/2006 06:45:46 PM:
>
> > > I don't think you can rename something that's already been renamed, at
> > > least not in the current implementation.
> >
> > Oh, you can't (I've tried before). I meant "would" in the hopeful
> > sense...
>
> Have you tried trunk? I thought you could now copy and rename stuff
> "infinitely" now.
In a working copy, yes, but not in the filesystem code.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
Posted by Mark Phippard <ma...@softlanding.com>.
Karl Fogel <kf...@google.com> wrote on 10/03/2006 06:45:46 PM:
> > I don't think you can rename something that's already been renamed, at
> > least not in the current implementation.
>
> Oh, you can't (I've tried before). I meant "would" in the hopeful
> sense...
Have you tried trunk? I thought you could now copy and rename stuff
"infinitely" now.
Mark
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
Posted by Karl Fogel <kf...@google.com>.
"Garrett Rooney" <ro...@electricjellyfish.net> writes:
> On 10/3/06, Karl Fogel <kf...@google.com> wrote:
>> "Garrett Rooney" <ro...@electricjellyfish.net> writes:
>> > That's assuming that those are intended to be sequential, I was
>> > assuming that they were intended to be simultaneous (i.e. swap the
>> > locations of A/B and X/Y), which we can't actually do at this point.
>>
>> Oh! Good point. Although if you put the classic tmp var in the
>> middle, would it work?
>>
>> mv A/B TMP
>> mv X/Y A/B
>> mv TMP X/Y
>>
>> If that all happens in one txn, would there be any record of TMP ever
>> existing? Would X/Y's history simply say it came from "A/B"@PREV_REV ?
>
> I don't think you can rename something that's already been renamed, at
> least not in the current implementation.
Oh, you can't (I've tried before). I meant "would" in the hopeful
sense...
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On 10/3/06, Karl Fogel <kf...@google.com> wrote:
> "Garrett Rooney" <ro...@electricjellyfish.net> writes:
> > That's assuming that those are intended to be sequential, I was
> > assuming that they were intended to be simultaneous (i.e. swap the
> > locations of A/B and X/Y), which we can't actually do at this point.
>
> Oh! Good point. Although if you put the classic tmp var in the
> middle, would it work?
>
> mv A/B TMP
> mv X/Y A/B
> mv TMP X/Y
>
> If that all happens in one txn, would there be any record of TMP ever
> existing? Would X/Y's history simply say it came from "A/B"@PREV_REV ?
I don't think you can rename something that's already been renamed, at
least not in the current implementation.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
Posted by Karl Fogel <kf...@google.com>.
"Garrett Rooney" <ro...@electricjellyfish.net> writes:
> That's assuming that those are intended to be sequential, I was
> assuming that they were intended to be simultaneous (i.e. swap the
> locations of A/B and X/Y), which we can't actually do at this point.
Oh! Good point. Although if you put the classic tmp var in the
middle, would it work?
mv A/B TMP
mv X/Y A/B
mv TMP X/Y
If that all happens in one txn, would there be any record of TMP ever
existing? Would X/Y's history simply say it came from "A/B"@PREV_REV ?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21745 - branches/fs-atomic-renames/notes
Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On 10/3/06, Karl Fogel <kf...@google.com> wrote:
> rooneg@tigris.org writes:
> > +What's wrong with this? Well, what if we need to rename A/B to X/Y so that
> > +we can later rename X/Y to A/B? Not sure if there's a work around for that,
> > +but I'm also not sure if there's a way to get that into the system anyway.
>
> I think when something like that happens in a single commit txn, we'd
> just "fold it down" (or "reduce to lowest common denominator", to use
> one more metaphor before actually saying exactly what I mean).
>
> In other words,
>
> mv A/B X/Y
> mv X/Y A/B
>
> should reduce to a no-op. Do we actually do this folding right now?
> I'm not sure; I think we should, though.
That's assuming that those are intended to be sequential, I was
assuming that they were intended to be simultaneous (i.e. swap the
locations of A/B and X/Y), which we can't actually do at this point.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org